Sie sind auf Seite 1von 2

Little Picard Theorem

S. Kumaresan Dept. of Math. & Stat. University of Hyderabad Hyderabad 500 046 kumaresa@gmail.com (Notes by Ajit Kumar)
Theorem 1 (Little Picards Theorem). Any entire non constant function on C can miss at the most one point of C. Proof of this theorem is a beautiful application of the Uniformization Theorem and various other results from Complex analysis and Algebraic Topology. This shows how various results from dierent branches come together to give a simple proof of non trivial results. First of all let us start with the Uniformization theorem. We will not prove this theorem. The proof is quite technical and involved. For a proof, we refer the reader to books on Riemann surfaces. Theorem 2 (Uniformization Theorem). Let X be a simply connected Riemann Surface. Then X is biholomorphic to precisely one of the following spaces: 1) P1 (C) 2) C 3) B(0, 1), the open unit disk in C. Proof. Let f H(C) miss two points, say, p and q. That is, f : C C \ {p, q} is onto. Since C \ {p, q} is open in C, it is a Riemann surface. This is not simply connected because we can nd a loop which winds p but not null homotopic. Hence there exists a simply connected Riemann surface X which is a universal cover for X. Let p : X C \ {p, q} be the covering map. Then there exists a lift of f . That is, there exists a holomorphic map g : C X such that p g = f . Let be the set of all deck transformations of this covering map p. Then act on X totally properly discontinuously and hence the quotient of X by is C \ {p, q}. By the Uniformization theorem X is either P1 (C), C or B(0, 1). We claim that X is B(0, 1). Granting the claim for a moment, we complete the proof. The lift g of f is a constant by Liouvilles theorem. Hence f = pg is constant. Thus if an entire function misses two points, then it is a constant. 1

We now prove the claim by contradiction. If X = B(0, 1), then either X = P1 (C) or X = C. We show that both these are untenable. Let, if possible, X = P1 (C). Since P1 (C) is compact, p(X) = C\{p, q} is compact, an absurdity. Assume that X = C. Since is the group of deck transformation, it is discrete. The set of biholomorphic automorphism of C is the group G = {fa,b : a C , b C}, where fa,b (z) = az + b. The group of deck transformation is a discrete subgroup of G. It is well-known that the only (closed) discrete subgroups of Rn are of the form Z1 . . . Zr , 1 , . . . r Rn are linearly independent over R. (Exercise.) In this case the only possibilities are Z or Z Z. Therefore, either (i) X/ C/Z C\{p, q} or (ii) X/ C/(ZZ) C\{p, q}. In the rst case, C/Z is homeomorphic to a cylinder. Hence its fundamental group is Z whereas that of C \ {p, q} is a free group on two generators. In the second case, C/(Z Z) is homeomorphic to a torus and hence is compact whereas C \ {p, q} is not. We would like to comment on the classical proof. One uses the so-called modular function h in that approach. The modular function actually gives a covering map of the upper half-plane H := {z C : Im z > 0} onto C\{0, 1}. Now if f : C C\{0, 1} is holomorphic, we get a lift g : C H of f : f = h g. Since H and B(0, 1) are biholomorphic, say, via , we see that the holomorphic function g is a constant by Liouvilles theorem. Hence g = 1 ( g) is a constant. Consequently, f = h g is a constant. If f : C C \ {p, q}, then we use the fact C \ {p, q} is biholomorphic to C \ {0, 1} to reduce it to the earlier case.